When a friend in Rome beats AI

Then I wrote to a good friend who lives in the Eternal City instead.

Beyond the joy of reconnecting, he came up with an easier and far less expensive route. That small choice says something I keep returning to: human connections remain essential precisely where AI performance is still not as “intelligent” as we expect. It can read our needs through prompts. But it has no way of knowing us well enough to anticipate what we actually think. Not yet.

Arriving at the conference, I faced a fundamental question: How do we remain human when technology is transforming everything and making it “better?”

That question was at the heart of the keynote delivered by Msgr. Lucio Adrián Ruiz, secretary of the Dicastery for Communication, at the Vatican Media Headquarters in Palazzo Pio, one of the most insightful sessions of the conference, and for me, one of the most honest formulations of what we are actually facing.

For the past two years, I have been following the ethical and anthropological discussions around AI, while in parallel educating myself on the mechanics behind large language model architecture, how these systems actually work, where they fall short, and what the gap between their performance and human intelligence means in practice. Not as a technologist. As a communicator trying to understand the tools my field is already deploying, often faster than we understand them.

What I keep seeing, across small organizations and large ones, is a consistent pattern: a focus on restructuring workflows to incorporate AI tools, without a serious assessment of the cultural changes that follow when we begin replacing human judgment with algorithmic methods in how we communicate. The conversation usually stays at the level of productivity, what we save in time, what we gain in output, and what it costs to deploy. Those are legitimate questions. But they are not the first ones.

When organisations communicate, to inform, to support, to connect, the strength of any of those outputs is proportional to how much humanity the reader can sense behind them. This is not a romantic claim about the irreplaceable magic of human writing. It is something more structural. What institutions offer, at their core, is not information. It is relationship. And relationship requires the perception of a presence, someone who has thought about your specific situation, who cares and knows something about you that no prompt can carry.

There is a reason the gap in reach between an institutional post and a grassroots story on social media is as wide as it often is. In my experience, the numbers are not always there because one is better written. They are there because they connect with people.

At ECIC, I encountered two different approaches to AI among religious communities.

Roman Catholics, coherently drawing on Pope Leo XIV’s encyclical Magnifica humanitas, are building a framework that resists the pressure to adopt first and reflect later. The encyclical grounds the technological moment in theological anthropology, asking what it means for the human person before asking what AI can do. Protestant churches gathered at ECIC, on the other hand, are focusing on practical ways to use AI tools to expand and strengthen their mission across Europe. Less doctrine, more discernment in practice.

Both approaches are working from something real. And both are, in different registers, asking the same underlying question: what does it mean to serve people well when the tools we use are changing faster than our understanding of them?

I don’t think one is right and the other wrong. I think we need the friction between them. And I think the communities doing this work at the intersection of faith, communication, and technology are exactly the ones who should be leading this conversation, not following it.

The deepest thing I took from the conference was not a keynote argument. It was a feeling confirmed again and again in the conversations across three days in Rome: that the communicators doing this work are, at their best, still genuinely trying to understand people. Not audiences. Not segments. People.

That distinction is precisely what algorithmic communication, however well calibrated, however plausibly human it sounds, cannot replicate. Not because it lacks data, but because it lacks accountability. It does not bear the weight of what it says the way a person does. It cannot carry a relationship forward beyond the output.

The success of this transition depends on whether we design systems that do not simply chase better performance, but actively protect connection and collaboration. Systems built by people who have thought carefully about what they want to keep human, and who are willing to resist the pressure to give it up in exchange for scale.

My friend in Rome gave me better directions than the algorithm. He also asked how I was doing, and we met later and shared time together. For now, that is the difference that still matters. Whether it remains so depends on the choices we are making right now.